Validation of the polyvoids in the design of bituminous mixtures with coal tar as a binder Validación del polígono de vacíos en el diseño de mezclas bituminosas con alquitrán de hulla como ligante

The world has developed different methodologies for the design of asphalt mixtures; each of which is intended to optimise and find the optimal combination of materials that allow them to be resistant under specific travel, climate and structural support conditions. This article presents the results of an investigation which utilised coal tar as a binding agent for road surface mixtures and carried out a comparison of results of the physical and mechanical properties of sample bituminous mixtures produced by two distinct methodologies, Marshall and Ramcodes. Finally, an analysis was carried out on the suitability of tar as a binding agent and on utilising the Ramcodes methodology when designing mixtures.


Introduction
Since the beginning of time, man has endeavoured to develop activities to allow survival and evolution.In principle, these activities were limited to the direct consumption of resources that nature provided, with the improvement of industrial activities increasing the presence of residues that were no longer bearable for the natural environment.
Road construction is an activity that implies a significant consumption of natural resources, during which the production of bituminous mixtures involves using great quantities of aggregate and asphalt; the majority of these coming from quarrying and oil exploitation.Taking into account the politics of sustainability, this investigation aims to contribute towards the development in the use of by-products in the iron and steel industry.
Furthermore, with comparisons to the Marshall methodology, this investigation aims to analyse the benefits of using the Ramcodes methodology in the design of bituminous mixtures and which out of these is most used in our environment.
In order to verify previous findings, with help from the SPSS programme, statistical analysis was performed with the results obtained by the two methodologies, to determine whether these results differ and whether the use of this methodology is advisable.

Materials used 2.1 Tar as a binding agent
Tar, as shown in Figure 1, is a semi-solid or liquid bituminous product that is obtained as a residue of distillation, in the absence of air, organic substances that contain volatile materials such as coal or wood.It is a cold binding agent, resistant to water, insoluble in lubricating oils and petroleum fuels, soluble in ether, benzene, carbon disulphide, chloroform and quinoline.The tar consists of 1 Corresponding Author: Professor from the Transportation and Public Works Department and active Investigator for GRINFRAVIAL.Tunja, Colombia.E-mail: ricardo.ochoa@uptc.edu.coaliphatic and naphthenic components containing hydrogen, oxygen, nitrogen and sulphur.
The tar utilised in this study is produced in a Coking plant of the iron and steel company Acerías Paz del Río S.A., whose carried out the process at 1000°C (Gómez, 2002).Table 1 shows the physicochemical properties of this material.

Aggregate
The materials utilised in a bituminous mixture must be subjected to rigorous studies to establish the possibility of their use.For this, since coarse aggregate uses a crushed type of gravel, this was subjected to the following tests: erosion in the Los Angeles abrasion testing machine, percentage of fractures, extension rate and flattening rate.
In addition, tests were carried out to determine the specific gravity and the percentage of absorption.As fine aggregate used sand and, as a filler mineral, it used Portland cement and grains of sand passed through sieve #200, their specific gravities were determined.All the tests were made taking into account the testing standards for road materials set by INVÍAS and ASTM.
The type of mixture selected whilst carrying out this study was a dense mixture in type 19 heat  in accordance with Article 450-13, of the general road construction specifications by INVÍAS (2013a).

Methodology
The methodology used in the development of this investigation was considered in three stages: the first was conducting tests in order to know the characteristics of the materials used in the design of the bituminous mixtures; the second stage regards carrying out necessary tests to obtain the working formula of the bituminous mixture for the Marshall methodology modified by University of Illinois, and for the Ramcodes methodology, and the third, with regard to the comparison of the results from both methodologies and the analysis in the advantage of using tar as a binding agent and the implementation of the Ramcodes methodology.

Marshall Method
The purpose of this method is to determine the optimum binding content for a specific aggregate mixture and to provide information on the physical and mechanical properties of hot bituminous mixture, so that it is possible to establish if they meet the parameters of densities and optimal content vacuum during the construction of the pavement layer.
The overall objective of the design process involves determining a combination and economic intensity of aggregate and binding agents that produce a mixture with: sufficient binding agent to guarantee a durable pavement; adequate stability to meet traffic demands without causing any deformation or displacement; a vacuum content high enough to allow a slight amount of additional compaction under heavy, without any bleeding or loss of stability and  enough workability to allow for efficient placement without segregation (Hosseinzadeh et al., 2016).

Ramcodes Method
The Ramcodes method (Ochoa-Díaz, 2013;Sánchez et al., 2002), an acronym for RAtional Methodology for COmpacted geomaterial's DEnsification and Strength analysis developed by F.J. Sánchez-Leal, is a methodology based on factorial experiments and the practical experience in design and control, for an analysis of densification and compacted geomaterial resistance.
The Marshall accelerated by Ramcodes is an application developed for the quick design of traditional Marshall procedure, which rationally binds the design specifications with the field control criteria, by implementing the "polyvoids" which defines an area which meets all vacuum specifications (VAM, VAF y VA) (Sánchez-Leal et al., 2011).These vacuums are based according to the binding content (%CA) and Bulk density (Gmb) and are indicated on maps as isolines, for the allowed values, the intersection of these lines produce a graphic construction of space %CA -Gmb, which results in a polygon, in which, through its centroid, is mathematically possible to obtain an asphalt content that meets all vacuum specifications for the mixture (Delgado et al., 2006).
The following steps have been proposed to modify or accelerate the original Marshall design process (Sánchez-Leal, 2009): Determine the specific effective gravity (Gse), the specific Bulk gravity of a combination of aggregates (Gsb) and the apparent specific gravity (Gsa) of the combination of selected aggregates.Verify that Gsa>Gse>Gsb, according to theoretical definitions.

Experimental design
In Table 2, six granulometric distributions of the material use are shown, with their respective dosages (Sánchez-Leal, 2007); the particle sizes were adjusted so that, in a strip of the MDC-19 mixture, they were exactly in the centre, on the thin side and on the thick side of this strip, Figure 2, in order to evaluate the impact of the particle size in the mixture's performance.
The variable is the type of mineral filler (cement or sand), adjusted to the granulometric strip of the mixture.The laboratory testing process for the Marshall and Ramcodes methodologies was developed in two stages.
In the first stage, six Marshall designs were carried out for experimental development, three with cement as a filler and three with sand passed through sieve #200.The heating temperate of the aggregates and the binding agent was 40°C with a constant compaction of 75 shocks per side (traffic level NT-2) (INVIAS, 2013b), making a factorial experiment varying in two factors, the binding content variable by 0.5% for each mixture and variable gradation according to the position on the granulometric strip.The former, with the objective of finding the working formula for the Marshall design, followed by the compaction process and outcomes of the mechanical performance of the mixtures.In the second part, six Ramcodes were developed (identical particle sizes to those used with Marshall), with constant binding percentages for each particle size, with a constant compaction of 75 shocks per side and a mixing temperature of 40°C.These percentages were obtained according to the use of the RAMSOFT program based on the Ramcodes methodology, which generated the work formula (optimum binding percentage) for each mixture, followed by the compaction process and outcomes of the mixtures stability and flow.Once the samples were compressed, they were left to cool for 15 minutes, before removing the mould.Once out of the mould, they were allowed to cool and cure at room temperature for eight days, so that the mixtures solvents evaporated.Following this, the test was continued to determine the stability and flow in the Marshall plan.

Results of the tests carried out with both methodologies
The following Table 3 and Table 4 present the results obtained from the mixtures designed through both methodologies.
Regarding the obtained results and taking into account the specifications of Table 450.4from article 450-07 of the general specifications of the construction of highways, the mixtures prepared with Portland cement as a filler achieve a level of transit NT-1; However, the flow is above the range of 2-4mm.
In the observed results of the mixtures prepared with sand passing the sieve #200 as a filler, some parameters achieve the specifications for a level of transit NT-1, although the flow resulted remotely from the specified range.The results of the Marshall Method for the Mixture M-2 fulfil all the specifications for the level of transit NT-2, to which this work is the formula that best presents mechanical behaviour.

Sensibility Analysis for Mixture M-2
For this mixture, the study of sensibility was achieved based on statistical techniques of the analysis of factorial experiments according to Ramcodes.Furthermore, the behaviour of the two most influential factors in bituminous mixtures (G mb y %P b ) was studied.This analysis has great advantages in controlling the compact mixtures.
Besides the vacant requisites, the bituminous mixture has to achieve the mechanical property's requirements such as stability and flow.The graphic representation of the response values under G mb and %P b generate a contour plot that is an image of the response surface for each selected mechanical parameter.
Ramcodes are based on a two-level factorial experiment-the binding content (%P b ) and the specific bulk weight (G mb ).The usage of maps permits a vision of variation of the mechanical properties inside the area where the space specifications are achieved.In Figure 3, the behaviour of stability is shown, while in Figure 4, the behaviour of flow is demonstrated.This mixture has high or low performance depending on where status is established.

Statistical Analysis
In order to do the comparison stability in relation to the used methodologies, the box diagram, Figure 5, is initially verified.It shows a small difference between the values of stability and the variability of similar data.
In order to confirm if a significant difference exists between the results of stability for the two methodologies, it is necessary to verify the assumption of normality, for which the results of the test of normality are shown in Table 5.
In a diagram or graph Q-Q Normal, every value observed (Y i ) is compared to the typical score NZ i that theoretically corresponded to that value in the normal standardised distribution.On the X-axis, the values observed are represented and ordered from least to greatest (stability and flow); in such an order, the typical, normal scores are represented (NZ i ).When a sample precedes a normal population, the corresponding points for each part are clustered around the diagonal represented in the diagram.The diversions in the diagonal indicate deviations in normality.
Figure 6 shows the diagram Q-Q Normal for the data established and obtained by the Marshall Method.With this information, the statistical inference developed relating the difference in the average value of stability for both methodologies.The results obtained in the SPSS software for this test is presented in Table 6.
With the assumption of variances, these results lead to the conclusion that no significant difference exists in the values of stability (p-value = 0.856>0.05).This implies that the use of any method does not generate a significant difference in the values of stability.
To compare the flow in respect to the two applied methodologies, the box diagram in Figure 8 is initially verified.Figure 8 shows a similar behaviour for the data of flow for both methodologies.
Using the Mann-Whitney Test (Table 7) for the independents samples, that which is observed in Figure 8 can be confirmed.
Regarding the confidence level of 95%, p-value =0.299>0.05,no significant difference between the values of flow can be concluded by using either methodologies.

Conclusions
It is not intended to replace asphalt for pavement in the preparation of mixtures.Instead, the intention is to seek alternatives that permit the selection of product that adjust to the needs of specific project.
Ramcodes is a methodology of great usefulness for its design, production and quality control of bituminous mixtures, provided that it guarantees the compliance of the volumetric parameters demanded by the specifications.Furthermore, there exists various benefits to using this methodology as it saves times, resources and money.Thankfully, it is only necessary to make three briquettes in comparison to 15 briquettes in the Marshall Method.
Amongst the inconsistent observations in the development of this investigation and keeping the statistical analysis in mind, it was demonstrated that no significant difference exists between the results obtained by the Marshall Method and the Ramcodes Method-being that the usage of the Ramcodes Method is valid in the design of the bituminous mixtures of pavements.
The mixture that best presents the behaviour and best adjusts to the required technical requests by INVIAS norms was the preparation with gravel as a course aggregate, sand as a fine aggregate, and cement as a filler.However, one must continue with this investigation in order to find the best way to utilise this bind and thereby contribute to the environment.
Given the thermal susceptibility of bitumen and, therefore, of the asphalt mix, it is determined that this type of mixture can be used in cold to temperate climates.Under these conditions, the behaviour of the mixture will be acceptable under the action of burdens imposed by traffic.
If some mixtures did not fulfil the INVIAS requirements for the wearing layer of the roadway, it would be possible to use them as enhanced base or sub-base layers.
To guarantee an acceptable behaviour of this type of mixture requires a strict control of quality in the minimum

Figure 2 .
Figure 2. The Granulometric Distribution of the Mixtures Figure 7 presents the diagram Q-Q Normal for the data established and obtained by RAMCODES method.The above information verifies that the normal distribution adequately traces the data of stability by methodology.Furthermore, the Shapiro-Wilk Test in Table 5 confirms this behaviour at a level of confidence of 95% (α=0.05)Marshall Method: p-value = 0.532>0.05RAMCODES Method: p-value=0.101>0.05

Figure
Figure 8. Box Diagram regarding Flow

Table 2 .
Mixture particle distributions and aggregate type

Table 3 .
Results of the Designed Mixtures with Cement as a Filler

Table 4 .
Results of the Designed Mixtures with Sand as a Filler

Table 5 .
Normality Test Results a. Correction of the significance of Lilliefors *.This is an inferior limit of true significance Figure 6.Diagram Q-Q Normal, Stability regarding the Marshall Method Figure 7. Diagram Q-Q Normal, Stability regarding Ramcodes Method Levene'

Table 6 .
Test Results of Independent Samples regarding Stability

Table 7 .
Results of the Mann-Whitney Test a.No corrections for the ties.b.Group Variable: Method